Barnett HJM, Gunton RW, Eliasziw M, Fleming L, Sharpe B, Gates P, Meldrum H. Causes and Severity of Ischemic Stroke in Patients With Internal Carotid Artery Stenosis. JAMA. 2000;283(11):1429-1436. doi:10.1001/jama.283.11.1429
Author Affiliations: The John P. Robarts Research Institute (Drs Barnett and Eliasziw and Mss Fleming, Sharpe, and Meldrum) and the Departments of Clinical Neurological Sciences (Drs Barnett and Eliasziw), Medicine (Dr Gunton), and Epidemiology and Biostatistics (Dr Eliasziw), University of Western Ontario, London; and the Department of Neuroscience, Geelong Hospital, Geelong; and the Department of Medicine, Melbourne University, Melbourne, Victoria, Australia (Dr Gates).
Context Therapeutic trials generally have not distinguished outcomes of stroke
according to cause.
Objective To determine whether stroke and subsequent disability was of large-artery,
lacunar, or cardioembolic origin in patients with different degrees of symptomatic
and asymptomatic carotid stenosis.
Design Observational study of prospective data collected from the North American
Symptomatic Carotid Endarterectomy Trial between 1987 and 1997.
Setting and Patients A total of 2885 patients from 106 sites in the United States and abroad
(median age, 67 years; 70% male) who had symptomatic internal carotid artery
Main Outcome Measure Risk of stroke from each of the 3 causes at 5 years by territory and
degree of stenosis.
Results During an average follow-up of 5 years, 749 patients had 1039 strokes,
including 112 of cardioembolic, 211 of lacunar, 698 of large-artery, 17 of
primary intracerebral hemorrhage, and 1 of subarachnoid hemorrhage origin.
The 5-year risk of first stroke after entry into the trial in any territory
was 2.6% of cardioembolic cause, 6.9% of lacunar cause, and 19.7% of large-artery
cause. The proportion of cardioembolic strokes in the territory of the symptomatic
artery was 12.0% and 6.9% in 60% to 69% and 70% to 99% arterial stenosis,
respectively; large-artery strokes predominated (78.4%) at 70% to 99% arterial
stenosis. With 70% to 99% arterial stenosis, the proportion of strokes of
cardioembolic and lacunar origin was 43.5% and 21.6% in asymptomatic and symptomatic
arteries, respectively. A total of 67.6% of cardioembolic, 16.7% of lacunar,
and 33.0% of large-artery strokes in the territory of the asymptomatic artery
were disabling or fatal.
Conclusions Our data suggest that approximately 20% and 45% of strokes in the territory
of symptomatic and asymptomatic carotid arteries with 70% to 99% stenosis,
respectively, are unrelated to carotid stenosis. The cause of subsequent strokes
in similar types of patients should be considered when making treatment decisions
involving carotid endarterectomy for patients with asymptomatic carotid stenosis,
since lacunar and cardioembolic strokes cannot be prevented by endarterectomy.
Observational studies and clinical trials generally have made little
attempt to distinguish between stroke outcomes according to cause,1,2 but this pattern is changing.3- 5 From the source documents
of the North American Symptomatic Carotid Endarterectomy Trial (NASCET), all
ischemic stroke outcomes were identified by cause (large artery, lacunar,
and cardioembolic). The working hypothesis was that a stenosing arterial lesion
should not be assumed to be the cause of all subsequent strokes originating
from this site. In the age group at risk of arteriosclerotic stenosis, concomitant
disease develops in other sites, including the heart and intracranial small
Identifying the causes of stroke among patients with carotid stenosis
is certain to assist physicians and patients making decisions about the acceptability
of a treatment program. Accordingly, the aims of this study were to determine
(1) the risk and the proportion of large-artery, lacunar, and cardioembolic
stroke in the territories of symptomatic and asymptomatic arteries; (2) the
occurrence of the different causes of stroke in patients with varying degrees
of carotid stenosis; (3) the influence of endarterectomy on the occurrence
of different causes of stroke; and (4) the level of disability of stroke related
The present study reports data gathered from the long-term follow-up
of 2885 patients with symptomatic carotid artery stenosis, half of whom received
carotid endarterectomy and half of whom received medical care alone.
NASCET confirmed the efficacy of carotid endarterectomy in reducing
the risk of stroke and death in patients who experienced a transient ischemic
event or nondisabling stroke and who had 70% to 99% stenosis of the ipsilateral
carotid artery.6 The benefit was reduced for
those with 50% to 69% stenosis.7 Patients with
less than 50% stenosis did not benefit. Patients in NASCET differed from other
patients with stroke in several large registries4,8- 12
and reviews.13,14 Individuals
with cardiac conditions likely to produce cardioembolism, including atrial
fibrillation (AF), recent myocardial infarction (MI), unstable angina, dilated
cardiomyopathy, recent congestive heart failure, and valvular disease prone
to the production of emboli, were excluded from the trial.
In total, 2885 patients with symptomatic internal carotid artery stenosis
were randomly assigned from December 1987 to December 1996 to undergo endarterectomy
or receive medical management alone. The average follow-up was 5 years. In
patients with hypertension, diabetes mellitus, and hyperlipidemia, best contemporary
treatment was administered and centrally monitored. Cigarette smoking was
discouraged. Enteric-coated aspirin was recommended for all medical and surgical
patients throughout the trial. In the event of recurrent ischemic events despite
aspirin use, other platelet inhibitors or anticoagulants were permitted at
the discretion of the treating physician.
Initial assessment included standardized history taking, history taking
and examination for cardiovascular symptoms or disorders, neurological examination,
laboratory tests, 12-lead electrocardiogram, chest radiograph, cross-sectional
imaging of the brain (usually computed tomography), angiography, and duplex
ultrasonography of the carotid arteries. Study neurologists in the participating
centers performed medical, neurological, and functional status assessments
on all patients at entry; at 1, 3, 6, 9, and 12 months; every 4 months thereafter;
and at the time of any stroke. Cardiology consultations were requested by
treating physicians when a cardioembolic origin of stroke was suspected.
All strokes and deaths were assessed at 4 levels: by the participating
neurologist and surgeon at each center, by staff neurologists at the data
center, by the outcomes committee (consisting of an internist, neurologists,
surgeons, and methodologists), and by external adjudicators. Among all strokes,
a distinction in territory was made among those on the same side as the recently
symptomatic randomized artery (ipsilateral), those on the opposite side (contralateral),
and those in the vertebrobasilar circulation. The severity (using a modified
Rankin scale at 90 days to determine stroke disability) and type of stroke
(hemorrhagic or ischemic) were assigned. A stroke was considered to be disabling
if the Rankin score was 3 or greater.
All ischemic strokes, including those within 30 days of endarterectomy,
were designated as to cause: large artery, lacunar, or cardioembolic. Outcomes
were evaluated for other potential uncommon stroke causes. Each stroke was
categorized as having 1 etiology only. Strokes not clearly lacunar or cardioembolic
in origin were categorized as large-artery strokes. Because all patients entered
the trial with symptoms or signs associated with a putative ischemic event
of large-artery origin, no strokes were expected to be assigned to an "unknown"
cause, and none were.
Lacunar strokes were defined by a combination of clinical and radiological
criteria: patients presenting with primary motor, primary sensory, or sensory-motor
symptoms; the dysarthria–clumsy hand syndrome or the ataxia-hemiparesis
syndrome; and with and without radiologically deep white matter lesions or
basal ganglia lesions no more than 1 cm in diameter.15
Cardioembolic strokes were identified by a combination of clinical and
echocardiographic features. Single or combined cardiac risk factors considered
to be indicative of a cardiac source for an embolic event were tabulated (Table 1). The guidelines for this classification
were consistent with several reports.13,16- 20,22- 24,30,31
Each cardioembolic stroke was designated as possible or probable. Criteria
influencing these 2 assignments were the number of associated risk factors
(in AF), the time relationship (in MI and invasive cardiac procedures), the
location of the cardiac abnormality (in MI and intracavitary thrombus), and
the characteristics of the associated clinical condition (in cardiac valvular
In nonvalvular AF (NVAF), the event rates for thromboembolism, derived
from multiple clinical and echocardiographic risk factors, provided the basis
for postulating a possible or probable cardiac source (Figure 1).16- 18
Strokes in patients with NVAF who had 2 cardiac risk factors were designated
as possible cardioembolic. These patients are predicted to have an annual
cerebral thromboembolism event rate of 11% to 12%, based on the sum of event
rates for 2 risk factors.16- 18
This is approximately 8 times higher than for NVAF patients without cardiac
risk factors (1.5% rate).18 When 3 or more
cardiac risk factors were identified, the stroke was designated as probable
cardioembolic. Nonvalvular AF in the presence of 3 or more risk factors carries
a considerably greater risk of thromboembolism (18%-37%) based on the sum
of event rates for 3 or more risk factors.16- 18
This is compared with the 4% annual rate for large-artery stroke observed
in symptomatic NASCET patients and the 2% annual rate observed in Asymptomatic
Carotid Atherosclerosis Study patients.32 Because
of such a marked discrepancy in the risk of thromboembolism between high-risk
NVAF and large-artery lesions, it was appropriate to select AF as the cause
of the stroke. No distinction was made between paroxysmal and established
AF.18 Patients with lone AF were not classified
as having a cardioembolic stroke.
When a cardioembolic cause appeared likely and required further clarification,
transthoracic echocardiograms (TTEs) were requested by the outcomes committee
if these studies had not already been obtained by the attending neurologist
or the consultant cardiologist. Transthoracic echocardiograms were not always
requested when AF developed or an acute Q-wave electrocardiogram was present
in the setting of clinical findings and enzyme changes consistent with MI.
Transesophageal echocardiography, refined in the closing years of the study,
was available for only a few patients. Left ventricular dysfunction was described
qualitatively as regional/local akinesia or dyskinesia or global dysfunction.
Left ventricular function was assessed by TTE at participating centers by
echocardiographers who reported on left ventricular segment wall motion abnormalities,
global left ventricular dysfunction, and ejection fraction measurements. Left
ventricular hypertrophy was based on the echocardiographic diagnosis.
Potential cardioembolic strokes identified by the outcomes committee
were separately adjudicated by the study cardiologist (R.W.G.), blinded to
patient treatment category and severity of carotid disease, using the template
in Table 1. Six months later,
the same cardiologist, again blinded, reviewed a random sample of 45 strokes,
deliberately including some already designated as noncardiac in origin. His
intraobserver κ was 0.96. A second cardiologist, also blinded, reviewed
these outcome events and the interobserver κ was 0.72.
The risk of each cause of stroke at 5 years in the medically treated
and surgically treated groups by territory (ipsilateral, contralateral, or
vertebrobasilar) and by degrees of carotid artery stenosis were estimated
from Kaplan-Meier event-free survival curves. For the purpose of correctly
ascribing strokes to medical or surgical therapy, the risk analyses censored
patients at the time of crossover to the alternate therapy from their assigned
treatment group at baseline. The risk analyses counted only the first stroke
of each cause in each territory.
The risk of each cause of stroke in the territory of an asymptomatic
carotid artery was also estimated in the subgroup of 1820 patients who had
an asymptomatic stenosed carotid artery on the contralateral side. The artery
was regarded as asymptomatic if there was no recent or remote history of symptoms
or physical signs in the territory of that artery. Presence of a lesion on
a computed tomography scan in the absence of symptoms or signs was considered
to be asymptomatic. The risk analyses censored data at the time of carotid
endarterectomy for individuals who subsequently had surgery on the side of
the asymptomatic artery.
In a further analysis, the risks of disabling and nondisabling stroke
of each cause, ipsilateral to the symptomatic and asymptomatic carotid territories,
were estimated from Kaplan-Meier event-free survival curves.
Baseline characteristics of patients in NASCET were previously reported.6,7 In brief, the median age was 67 years
and 70% were men. There was a recorded history of hypertension (61%), diabetes
mellitus (22%), hyperlipidemia (34%), smoking within the past year (42%),
and prior MI or angina (36%). Spanning the duration of the trial, a total
of 1039 strokes (1021 ischemic, 17 primary intracerebral hemorrhage, and 1
subarachnoid hemorrhage) involving any territory occurred in 749 of 2885 patients
in the NASCET. Recurrent strokes in the same patient did not always originate
from the same cause. Cardioembolic origin accounted for 112 (10.8%) of the
strokes occurring in 95 patients, lacunar strokes were identified 211 times
(20.3%) in 184 patients, and large-artery disease strokes occurred 698 times
(67.2%) in 542 patients.
Among the 95 patients who had cardioembolic strokes, 80 had 1 stroke,
13 had 2 strokes, and 2 had 3 strokes. Sixty-two cardioembolic strokes were
designated possible and 50 probable. Echocardiograms were performed in 64
The most frequent cause was NVAF, identified in 39 of the 112 cardioembolic
strokes (34.8%) (Table 1). Twenty-two
strokes were assigned to the probable category and 17 to the possible category.
Of the 39 strokes occurring in 34 patients who developed AF, 5 patients were
taking anticoagulants prior to the event and 12 others began taking long-term
anticoagulants after the event. During follow-up, another 195 patients developed
NVAF without experiencing a stroke, 56 of whom were given anticoagulants.
Regional/global left ventricular dysfunction based on TTE findings at
the time of the outcome event was deemed to be due to ischemia and classified
as a possible cause in 37 (33%) of 112 strokes. Excluding left bundle-branch
block in 4 patients, 23 of the 33 remaining strokes were associated with significant
electrocardiographic Q waves recorded at the time of or prior to the outcome
event, strongly suggesting transmural MI. Three strokes involved patients
with previous coronary artery bypass graft; 20 (54%) had a remote history
Acute MI was associated with 11 cardioembolic strokes. In 6 patients
with acute anterior MI, a probable cause was assigned. The strokes occurred
within 3 weeks of the infarction. A possible cause was assigned in 5 other
patients when the stroke occurred within 3 months of the infarction.
Invasive cardiac procedures were identified as the probable cause of
17 strokes occurring within 24 hours of the procedure. Eleven followed cardiac
surgical procedures, coronary artery bypass graft, or aortic valve replacement,
and 6 followed catheterization procedures, coronary angiography, and/or percutaneous
Other cardiac risk factors for probable and possible strokes of cardioembolic
origin are listed in Table 1.
Of the 1021 ischemic strokes, 727 remained in the risk analyses after
accounting for crossovers, first occurrence, and the 5-year period. The 5-year
risk of a cardioembolic stroke in any territory was 2.6% (95% confidence interval
[CI], 2.0%-3.4%), approximately one third the risk of lacunar stroke (6.9%)
and one seventh the risk of large-artery stroke (19.7%) (Figure 2, top). Cardioembolic strokes in any territory occurred
equally in medically and surgically treated patients (2.6%). Patients were
at reduced risk of all causes of stroke in the contralateral and vertebrobasilar
The proportion of each cause of stroke was calculated using data from
all 1021 ischemic strokes. Thirty percent of all strokes were not of large-artery
origin, irrespective of treatment group and territory (Figure 2, bottom). Cardioembolic strokes occurring on the side of
the randomized artery (ipsilateral to symptoms) accounted for 8.5% and 9.9%
of all strokes in the medical and surgical groups, respectively.
In medically treated symptomatic patients, the risk of an ipsilateral
large-artery stroke increased as expected with severity of stenosis (Figure 3, top). The risk of lacunar and cardioembolic
strokes did not rise with increasing stenosis. In the territory of an asymptomatic
artery, there was a lower risk of all causes of stroke (Figure 3, top). The risk of large-artery stroke increased with the
degree of stenosis, but even in the 70% to 99% stenosis category, it only
reached about one third of that in the symptomatic artery. Cardioembolic strokes
did not increase in the presence of increasing stenosis in the asymptomatic
artery. The risk of lacunar strokes increased moderately.
For symptomatic arteries with severe (70%-99%) stenosis, the highest
proportion of strokes were of large-artery origin (78.4%) (Figure 3, bottom). Of all strokes in the territory of a symptomatic
70%- to 99%-stenosed artery, 21.6% were cardioembolic or lacunar in origin
(6.9% + 14.7%, respectively). At lesser degrees of stenosis, although large-artery
strokes still predominated, the proportion of lacunar and cardioembolic strokes
was higher than in the territory of a 70% to 99% stenosis.
For asymptomatic arteries, the proportion of strokes of combined lacunar
and cardioembolic origin accounted for one third to nearly half of strokes
(Figure 3, bottom). The proportion
of lacunar and cardioembolic strokes in the territory of the asymptomatic
artery was somewhat higher in subjects with greater degrees of stenosis. Of
all strokes on the asymptomatic 70%- to 99%-stenosed side, 43.5% (8.7% + 34.8%,
respectively) were of cardioembolic or lacunar origin.
The 5-year risks of disabling and nondisabling strokes in the territory
of a symptomatic carotid artery for medically treated and surgically treated
patients are shown in Figure 4,
top. The risk of stroke of any degree of disability was highest in medically
treated patients whose strokes were attributed to large-artery disease (11.0%
nondisabling and 5.6% disabling). The risks of disabling stroke due to lacunar
and cardioembolic causes were only 0.2% and 1.2%, respectively. The 5-year
risk of stroke by cause in the territory of asymptomatic arteries is shown
in Figure 4, top. The reduced risk
of large-artery stroke compared with the territory of a symptomatic artery
is shown, as are the equal risks of cardioembolic and lacunar strokes.
The proportion of ipsilateral stroke by cause and disability was calculated
(Figure 4, bottom). About 51% (25.8%
+ 25.8%) of the cardioembolic strokes in the territory of a symptomatic artery
in medically treated patients were disabling. The proportion of disabling
strokes of lacunar cause was substantially lower (4%). None were in the most
severe or fatal categories. The proportion of disabling or fatal large-artery
strokes was 33.7% (19.8% + 13.9%, respectively).
Similarly, in the territory of an asymptomatic artery, strokes that
were disabling or fatal were more common with cardioembolic cause (67.6%)
than disabling strokes of lacunar (16.7%) or large-artery (33.0%) cause (Figure 4, bottom). In the territory of an
asymptomatic artery, more lacunar strokes had no disability (Rankin score
of 0) at 90 days than was observed in strokes of cardiac or large-artery origin.
The results of this study demonstrate that even in the presence of large-artery
lesions, strokes arise from a variety of causes. To our knowledge, no previous
large trials or case-series of carotid endarterectomy have attempted to assign
the proportion of subsequent strokes by cause. This omission may be of critical
consequence, considering that subsequent strokes arising from cardiac or aortic
sources and the penetrating arteries of the brain (producing lacunes) are
not likely to be prevented by an operative procedure (endarterectomy) that
is not expected to eradicate these potential causes.
Although previous data bank and case-series studies have examined the
relative disability of strokes by cause,34- 39
the disability imposed by the various causes of stroke had not been addressed
in patients with proven carotid artery disease. The present study shows that
strokes of cardioembolic origin account for the highest proportion of disabling
strokes. Strokes of lacunar origin are least disabling, and those attributable
to large arteries have a disability between strokes of cardioembolic origin
and lacunar origin, whether the subsequent strokes were in the territory of
the symptomatic artery or the asymptomatic artery.
Cardiogenic stroke was diagnosed less often in this trial than in large
case-series and data banks. Stroke data banks have identified cardiac causes
in 15% to 20% of strokes,9- 11,40- 42
whereas in NASCET, the proportion was lower at 10.8%. This finding is attributable
to exclusion from entry in NASCET of patients with cardiac diseases known
to be a source of embolism. These patient-selection criteria imply that the
observations from the present study cannot be generalized to all patient populations
with carotid stenosis but are applicable to patients being considered for
carotid endarterectomy in whom serious cardiac conditions were judged to be
a barrier to the procedure. The proportion of cardioembolic events in this
study may only be half that occurring in patients commonly undergoing endarterectomy
in community practice.
Not surprisingly for this population of patients with a median age of
67 years, NVAF developing after randomization headed the list of probable
causes of cardioembolic stroke. Atrial fibrillation becomes more common with
increasing age, affects 2 million people in the United States in any given
year, causes 36% of strokes among patients older than 80 years, and increases
stroke risk 6-fold.43,44 Two large
stroke registries have identified AF as the most common risk factor for cardioembolic
stroke.10,12 Anticoagulant therapy
(for which benefit has been proven since NASCET began) was administered to
only about one quarter of these patients, but should be recommended more frequently
in the future.
The second most common risk factor identified was regional/global left
ventricular dysfunction. Both left atrial thrombi and left ventricular thrombi
as cerebral embolic sources have been associated with left ventricular wall
motion abnormalities23 and left ventricular
dysfunction,24,25 with and without
In addition to NVAF, cardiac sequelae of atherosclerosis developed in
NASCET subjects, including symptomatic coronary heart disease, MI, and congestive
heart failure, in some cases requiring invasive therapeutic interventions.
All these factors are known to have potential for the formation or dislodgement
of aortic, atrial, and ventricular thrombi or atheromatous fragments causing
Among ipsilateral strokes in symptomatic patients with 70% to 99% stenosis,
a small proportion (6.9%) were of cardioembolic origin. By contrast, patients
with moderate (60%-69%) symptomatic stenosis had 12.0% of their strokes from
cardiac sources. The proportion of lacunar and cardioembolic strokes (approximately
35%) in symptomatic patients with less than 70% stenosis was approximately
twice that observed in patients with 70% to 99% stenosis. This proportion
of strokes not caused by large-artery lesions must be weighed in any decisions
about endarterectomy for patients with only moderate stenosis.
Of strokes occurring in the territory of the asymptomatic artery, the
proportion of large-artery origin was less than in the territory of the symptomatic
artery, particularly in the presence of severe stenosis. For patients with
asymptomatic arteries with 60% to 99% stenosis (the range of stenosis believed
to benefit from endarterectomy), nearly half of strokes were from the small
vessels or the heart.
Disorders of the heart and small penetrating arteries are important
causes of stroke, even in the presence of stenosing arteriosclerotic extracranial
carotid lesions. Nonvalvular AF and the consequences of myocardial ischemia
account for most of the strokes of cardiac origin. The risk and proportion
of cardioembolic stroke outcome events in this group of patients with symptomatic
carotid artery disease was similar in the medically and surgically treated
patients. Surgical treatment of the diseased artery would not be expected
to reduce the later occurrence of cardioembolic stroke and did not. Strokes
of cardioembolic origin were more commonly disabling or fatal compared with
those that arose from the large or small arteries.
Approximately 35% of strokes occurring subsequently in symptomatic patients
with moderate carotid disease (<70% stenosis) were due to causes other
than the large-artery lesion. Approximately 50% of strokes occurring in the
territory of the asymptomatic artery with 60% to 99% stenosis were not of
Expectations of benefit from endarterectomy in symptomatic patients
with moderate disease or in asymptomatic subjects must consider the evidence
that a large number of future strokes will not be related to the carotid artery
lesion. The risk of stroke expected from the large-artery lesion is only part
of the story. In patients with carotid stenosis suspected of causing ischemic
events, risk factors must be modified. Before deciding on the final treatment
strategy, consideration should be given to investigations to identify other
potential causes (eg, cardiac, aortic, lacunar, prothrombotic hematological
conditions) of stroke.